Need Information

Need Information

[Y.G Kumar]

Hi,

I am trying to understand the concept of cpu steal time. I have a few
questions about it.

According to the link
https://scoutapm.com/blog/understanding-cpu-steal-time-when-should-you-be-worried
 , it is defined as follows:

"Steal time is the percentage of time a virtual CPU waits for a real
CPU while the hypervisor is servicing another virtual processor."

I have two questions regarding that.

1. What do we mean by "percentage of time" ? Which time is it alluding
to ? Is it the "cpu time" allocated by the scheduler to that kvm
process ? Or, is it the general waiting time that vcpu process has to
wait in the run queue before getting the cpu for execution ? If this
is true, then it is normal for that vcpu process like any other
process on the system to wait in the runqueue. Then what is this fuss
about ?

2. How can we read this steal time ? Can we see it from within the vm
and from the hypervisor host also ?  If so, how ?  Does this mean that
we cant see the "st" field if we execute the top command on a server
without any hypervisor ? Please explain.

Thanks
Kumar

Need information

[Y.G Kumar]

Hi All,

Is there a way to collect cpu steal time metrics of a kvm-qemu virtual
machine from the hypervisor on which the vm is running ? Please give
me some opinions ..

Thanks
Y.G